Neil Croft wrote in a message to Mike Bilow:
 MB> But no one needs gigabit Ethernet to the desktop.
 NC> Care to be reminded of that line in, say, 5 years time Mike?
 NC> Whether we like it or not, bright spark users will keep
 NC> having daft ideas about IP phones and video conferencing
 NC> from the desktop.
Since a number of people seem intrigued by my remark, let me take this a step 
further.  I would contend that, although people may someday want gigabit 
BANDWIDTH to the desktop, it will not use Ethernet.
Ethernet is a lousy protocol, and it falls apart at high speeds.  It is 
inherently half-duplex, it depends upon CSMA/CD which in turn depends upon 
the introduction of random time delays, and these time delays start to become 
significant in terms of phase velocity even at the speed of light.  If we 
need very high bandwidth, this will dictate abandoning CSMA/CD and, 
therefore, abandoning Ethernet.  (Yes, I know that 100VG-AnyLAN is not 
CSMA/CD, but it also is not Ethernet, either.)
 NC> The network we're replacing in our office
 NC> was a nice, routed, 5 ring 4meg T/R with about 80 users on
 NC> each ring. The new one looks like it's going to be a
 NC> switched 10meg Ethernet with 100meg server farm and 12 users
 NC> per hubbed segment and the 3174s on 16meg T/R with a then
 NC> gradual switch to 100meg Ethernet. Next will come an FDDI
 NC> backbone no doubt. Simple fact of life is that Neil's 4th
 NC> law of computers is...
 NC> 4. You can never have enough bandwidth.
 NC> which follows processor, memory and hard drive space. =o)
Considering your reference to 3174 boxes, I would have to question the wisdom 
of your current upgrade plans.  I certainly hope that IBM has improved their 
Ethernet capability on the 3174, because I was personally involved with a 
mess in 1995 which lasted about four months.  There was actually an IBM 
representative on-site full-time for about three months of this, working on 
nothing else.  After a few of those $500 per hour telephone conference calls 
with IBM networking in Raleigh, several upgrades to the 3174 monitor with 
debugging builds, and other horrendous things I don't even remember, it was 
finally made to work.
And, as you might expect, the original motivation for using Ethernet instead 
of SNA token ring was cost savings!  This is well in tune with Bilow's First 
Law of Mainframes: "All economy is false economy."
 
-- Mike
--- 
---------------